Sebastian J. Spicker

The smartphone camera is used for exoplanet transit photometry, the onboard sensors for orientation and alignment. The measurements are real observations, not simulations.

Students formulate the observation hypothesis, calibrate the instrument, and identify sources of error themselves. The data feeds into a dataset shared across the participating institutions.

The hardware is deliberately low-cost: students who build an instrument themselves can see what it measures and where its limits are.

Vocabulary incompleteness: the model’s world knowledge exceeds its output vocabulary. Logit lens analysis shows that the correct concept is represented internally, but no matching token exists in the fixed vocabulary, so the model approximates without signalling the gap.

Grounding failure: models have no access to the current state of the world. Asked whether to drive to a car wash in the rain, a model averages over its training contexts; the problem is not missing knowledge but missing access to the present.

Pragmatic inference failure: the model has all the information it needs but reads a question literally where cooperative intent (Grice’s maxims) would be required. Each failure type calls for a different mitigation.

Music students applied structured reflection protocols to their own artistic decisions and treated practice sessions as data. The research question: does making implicit processes explicit support artistic development?

Music and physics both combine formal structure (voice-leading rules, physical laws) with exploratory practice, so research methods transfer between the two fields.

The reflection formats developed in the project are now part of the teaching at the HfMT Cologne.

The lessons were designed from scratch rather than adapted from existing material, because adaptations tend to keep the hidden assumptions that create barriers.

Design principles: a fixed opening routine, a consistent visual layout, and advance notice of every transition. Pacing follows processing speed rather than curriculum coverage.

The modules were revised based on classroom observation: points where students disengaged were treated as problems of the design, not of the students.